Self- Organizing Map and Multidimensional Scaling in a Tandem Approach: a Visualization of Bankruptcy Trajectory

In this paper, two innovations to a trajectory approach using Self- Organizing Maps are introduced to analyze financial statements of U.S. listed companies between 2013-2018, approaching changes in these statements as movement patterns. The first innovation entails a projection of quarterly statements on multiple maps instead of a single Self- Organizing Map. The second entails the use of a tandem approach between the Self-Organizing Map and Multidimensional Scaling to allocate the neurons in a Self-Organizing Map a more optimal position on the grid. By applying Multidimensional Scaling to the neurons in a configured Self- Organizing Map the visual representation of this map portrays more intuitive insights into the neighboring relations between each neuron. Including both innovations in the trajectory approach resulted in a slightly more accurate prediction of bankruptcy compared to the traditional Self- Organizing Map approach. By its ability to capture both cross-sectional and longitudinal differences between companies and depicting these in a two-dimensional intuitive map, the SOM- MDS tandem is a highly relevant tool for anyone seeking support in their company viability analysis.